Implicational classes ofde Morgan Boolean algebras

Transcription

1 Discrete Mathematics 232 (2001) Implicational classes ofde Morgan Boolean algebras Alexej P. Pynko Department of Digital Automata Theory, V.M. Glushkov Institute of Cybernetics, National Academy of Sciences of Ukraine, Academik Glushkov Prosp. 40, Kiev , Ukraine Received 13 December 1998; revised 15 February 2000; accepted 17 April 2000 Abstract An abstract algebra A; ; ; ; is called a De Morgan Boolean algebra if A; ; ; ; is a Boolean algebra and A; ; ; is a De Morgan lattice. In this paper we prove that implicational classes of De Morgan Boolean algebras form a four-element chain and are all nitely-axiomatizable and nitely-generated quasivarieties, three ofwhich are varieties. We also show that there are exactly two (up to isomorphism) subdirectly irreducible De Morgan Boolean algebras. c 2001 Elsevier Science B.V. All rights reserved. MSC: primary: 06E25; secondary: 03G10; 06D30; 06E05; 08B15; 08B26; 08C15 Keywords: De Morgan Boolean algebra; Boolean algebra; De Morgan lattice; Implication; Implicational class; Quasi-identity; Quasivariety; Identity; Variety; Subdirectly irreducible algebra 1. Introduction The concept of De Morgan Boolean algebra has been introduced in [8] in connection with the expansion proposed therein by the classical negation ofthe four-valued logic originally motivated by certain applications to Computer Science and Articial Intelligence in [3] and then examined within the context ofgeneral and Algebraic Logic in [7 9]. As opposed to the underlying four-valued logic [3], its expansion by the classical negation includes, as its fragment, the classical logic and, moreover, is functionally complete (see [8] for more detail). In [8] it has also been proved that the variety ofde Morgan Boolean algebras is equivalent, in the sense of[9], to the Gentzen calculus for the mentioned expansion of BELNAP s four-valued logic which has been introduced in [8]. In this paper we study the lattices ofvarieties and implicational classes ofde Morgan Boolean algebras and subdirectly irreducible De Morgan Boolean algebras. It appears address: pynko@d105.icyb.kiev.ua (A.P. Pynko) X/01/$ - see front matter c 2001 Elsevier Science B.V. All rights reserved. PII: S X(00)

2 60 A.P. Pynko / Discrete Mathematics 232 (2001) that these issues are dierent from those for both De Morgan lattices and De Morgan algebras. It is also worth noticing that, by nding quasi-varieties (varieties) ofde Morgan Boolean algebras together with their relative axiomatizations, we immediately get, according to the general theory ofequivalence developed in [9], the lattice of (axiomatic) extensions, together with their relative axiomatizations, ofthe Gentzen calculus mentioned above. Thus, studying varieties and implicational classes ofde Morgan Boolean algebras is not only interesting within the context oflattice Theory but also important for studying the expansion of BELNAP s four-valued logic by the classical negation. The rest ofthe paper is as follows. In Section 2 we recall basic concepts and results concerning implicational classes, quasivarieties, varieties and subdirectly irreducibles. In Section 3 we nd and study all the varieties and implicational classes ofde Morgan Boolean algebras and subdirectly irreducible De Morgan Boolean algebras. 2. General background The reader is supposed to be familiar with basic concepts and results concerning Universal Algebra and Lattice Theory (especially the theory of distributive lattices), for which one can consult, e.g., [2,5]. Below we recall those algebraic issues which are key for our study. These are used tacitly throughout the rest of the paper. Given an algebraic signature, a -equation is an expression ofthe form where and are two -terms with arbitrary variables (not from any xed set). A -implication is a couple ofthe form, where is a set of -equations and is a single -equation. In case is nite, is referred to as a -quasi-identity. In case =, is referred to as a -identity and is identied with. Let be a -implication and V the set ofall variables occurring in it. Then is said to be satisedin a -algebra A provided, for every : V A, 1 A []= A [] whenever A []= A [] for each. A class of -algebras K is said to be axiomatizedby a system of -implications S (relatively to a class of -algebras K )ifk is the class ofall -algebras (in K ) which satisfy each implication in S. A class of -algebras is said to be implicational whenever it is axiomatized by some system of -implications. A class of -algebras K is implicational i it contains the direct product ofany indexed subset ofk and any algebra embeddable into (in particular, any subalgebra and any isomorphic copy of) any element of K. A class of -algebras is referred to as a quasivariety whenever it is axiomatized by some system of -quasi-identities. A class of -algebras is referred to as a variety whenever it is axiomatized by some system of -identities. 1 In order to unify notations, unless otherwise specied, algebras are denoted by Calligraphic letters (possibly with indices), their carriers being denoted by the corresponding capital Italic letters (with the same indices).

3 A.P. Pynko / Discrete Mathematics 232 (2001) Given an arbitrary class of -algebras K, the implicational class axiomatized by the system ofall -implications satised in each algebra in K is the least implicational class including K. It is denoted by I(K) and is said to be generatedby K. A-algebra A belongs to I(K) i, for all a; b A such that a b, there are some B K and some h hom(a; B) such that ha hb. The class ofall subalgebras ofmembers of K is denoted by S(K). Notice that every member of I(K) is isomorphic to a subdirect product ofmembers ofs(k). Proposition 2.1. Let K be a class of -algebras. Then; the subdirectly irreducibles of I(K) are exactly the isomorphic copies of subdirectly irreducible subalgebras of members of K. Moreover, a -identity is satised in a subdirect product ofan indexed set {A i } i I of -algebras i it is satised in each A i, i I. Therefore, we also have Proposition 2.2. Let K be a class of -algebras and I a set of -identities. Then; the implicational class axiomatizedby I relatively to I(K) is the implicational class generatedby the class axiomatizedby I relatively to S(K). Corollary 2.3. Let K be a class of -algebras such that I(K) is a variety. Then; every subvariety of I(K) is the implicational class generatedby some S S(K). Propositions 2.1, 2.2 and Corollary 2.3, being a part ofthe algebraic folklore, are rather useful tools for analyzing subdirectly irreducibles and subvarieties. 3. Main results Throughout this section, we deal with the signature = ; ; ; where ; are binary, ; are nullary are unary, and its two subsignatures B = ; ; ; and D = ; ;. For the sake ofbrevity, we write 4 for. A De Morgan Boolean algebra [8, Denition 3:19] is any -algebra whose B -reduct is a Boolean algebra [2] and whose D -reduct is a De Morgan lattice [6] (viz., a distributive i-lattice in the sense of[4]), that is, a D -algebra whose ; -reduct is a distributive lattice and that satises the following identities: x x; (1) (x y) x y; (2) (x y) x y: (3)

4 62 A.P. Pynko / Discrete Mathematics 232 (2001) Fig. 1. The diamond De Morgan Boolean algebra BD 4. The variety ofall De Morgan Boolean algebras is denoted by DMBA. Notice that any De Morgan Boolean algebra also satises the identities: ; ; x: (6) Moreover, the ; ; ; ; -reduct ofany De Morgan Boolean algebra is a De Morgan algebra in the sense of[2]. By BD 4 we denote the -algebra whose B -reduct is the four-element diamond Boolean algebra with carrier BD 4 :={0; ; ; 1} and ordering 6 given by 0 1 and 0 1 (see its Hasse diagram in Fig. 1) and whose operation is dened by 0:=1, :=, := and 1:=0. The set BD 2 :={0; 1} forms a subalgebra of BD 4 that we denote by BD 2. The D -reducts of BD 4 and BD 2 are denoted by D 4 and D 2, respectively. Proposition 3.1 (cf. [8, Theorem 3:20]). DMBA= I(BD 4 ). Proof: Clearly, BD 4 is a De Morgan Boolean algebra. Hence I(BD 4 ) DMBA: Conversely, consider any De Morgan Boolean algebra A. Take any a; b A such that a b. By [10, Proposition 3:2] there is then some h hom(a D ; D 4 ) such that ha hb. Then ha forms a non-singular subalgebra of D 4. Hence BD 2 ha, sobythe following lemma h hom(a; BD 4 ). Lemma 3.2. Let A; B DMBA and h hom(a D ; B D ). Assume { ; } ha. Then h hom(a; B). Proof: As { ; } ha; h = and h =. Moreover, for all a A, we have ha h@a = and ha h@a = for = and =. Therefore h@a so h hom(a; B). Thus A I(BD 4 ).

5 A.P. Pynko / Discrete Mathematics 232 (2001) Notice that BD 2 is the only proper subalgebra of BD 4. Thus, S(BD 4 )={BD 4 ; BD 2 }: (7) Moreover, both BD 4 and BD 2 are non-singular and simple, so they are subdirectly irreducible. Hence, by (7) and Propositions 2.1 and 3.1, we get Corollary 3.3. A De Morgan Boolean algebra is subdirectly irreducible i it is isomorphic to either BD 4 or BD 2. Next, the identity x is satised in BD 2 but is not satised in BD 4. Hence, by (7) and Propositions 2.2 and 3.1, we also get Corollary 3.4. I(BD 2 ) is the subvariety of DMBA relatively axiomatizedby the single identity (8). Notice that I( ) is the least implicational class and consists ofall singular -algebras, that is, it is the variety axiomatized by the single identity x y. On the other hand, BD 2 is not singular. Thus, by (7), Proposition 3.1 and Corollaries 3.4 and 2.3, we get Corollary 3.5. The varieties of De Morgan Boolean algebras form the three-element chain I( ) I(BD 2 ) DMBA: Remark that the lattice ofvarieties ofde Morgan Boolean algebras is dierent from those of De Morgan lattices and De Morgan algebras that form a four-element chain [4,2] (cf. [8, Corollary 4:11]). Likewise, as opposed to the varieties ofde Morgan lattices and De Morgan algebras that have exactly three (up to isomorphism) subdirectly irreducibles, the variety ofde Morgan Boolean algebras has just two (up to isomorphism) subdirectly irreducibles. Proposition 3.6. I(BD 4 BD 2 ) is the subquasivariety of DMBA relatively axiomatizedby the single quasi-identity x x x y: (9) Proof: As BD 2 is a subalgebra of BD 4, by Proposition 3.1 we see that BD 4 BD 2 DMBA: Moreover, BD 4 BD 2 satises the quasi-identity (9). Conversely, consider any De Morgan Boolean algebra A satisfying the quasi-identity (9). Take any a; b A such that a b. Then, by Proposition 3.1 there is some h hom(a; BD 4 ) such that ha hb. Moreover, as A D is a De Morgan lattice satisfying the quasi-identity (9), by [10, Proposition 4:2] there is some g hom(a D ; D 4 D 2 ). As the right projection p :BD 4 BD 2 BD 2 ; c; d d is a homomorphism from D 4 D 2 onto (8)

6 64 A.P. Pynko / Discrete Mathematics 232 (2001) D 2, f:=p g hom(a D ; D 2 ). Notice that D 2 has no proper subalgebra. Hence fa=bd 2, so by Lemma 3.2 we get f hom(a; BD 2 ). Then, the mapping h f : A BD 4 BD 2 ;c hc; fc is a homomorphism from A to BD 4 BD 2. Moreover, (h f)(a) (h f)(b). Thus A I(BD 4 BD 2 ). It appears that I(BD 4 BD 2 ) is the only implicational class ofde Morgan Boolean algebras that is not a variety, as it follows from Corollary 3.5 and Theorem 3.7. The implicational classes of De Morgan Boolean algebras form the four-element chain I( ) I(BD 2 ) I(BD 4 BD 2 ) DMBA: Proof: First ofall, notice that, given a non-singular De Morgan Boolean algebra A, by (4) and (5) the mapping e :BD 2 A, dened by e0:= ; e1:= is an embedding of BD 2 into A. Hence, any implicational class ofde Morgan Boolean algebras other than I( ) includes I(BD 2 ). Next, it is easy to see that BD 4 BD 2 does not satisfy (8) (for BD 4, which is a homomorphic image of BD 4 BD 2, does not). Moreover, BD 4 does not satisfy (9). Hence, by Propositions 3.1, 3.6 and Corollary 3.4 we get I( ) I(BD 2 ) I(BD 4 BD 2 ) DMBA: Finally, take any implicational class ofde Morgan Boolean algebras K distinct from both I( ) and I(BD 2 ). Then I(BD 2 ) K. Consider the following two cases: 1. K I(BD 4 BD 2 ). Take any A K\I(BD 2 ), in which case A is not singular. Then, by Corollary 3.4, there is some a A such that Put b:=a a. Consider the mapping e from BD 4 BD 2 to A dened by e 1; 1 := ; e ; 1 := b; e ; 1 :=@ b; e 0; 1 := b; e 0; 0 := ; e ; 0 := b; e ; 0 b; e 1; 0 := b: Taking into account that b6b and using (1) (6), it is straightforward to check that e hom(bd 4 BD 2 ; A). We are going to show that e is injective. First, we prove, by contradiction, that the four pairs {e c; 0 ;e c; 1 }, where c BD 4, are pairwise disjoint. For suppose {e c; 0 ; e c; 1 } {e d; 0 ; e d; 1 } for some c; d BD 4 such that c d. Then e c; 0 6e d; 1 and e d; 0 6e c; 1, soe c d; 0 6e c d; 1. Moreover, as c d, we have c d c d. Therefore, either c d and c d6, in which case e ; 0 6e ; 1, orc d and c d6, in which case e ; 0 6e ; 1. Applying to the last inequality, we get e ; 0 6e ; 1 too. Thus, in any case, we have e ; 0 6e ; 1, that so a a = b =. Applying to the last equality, we also get a a =.

7 A.P. Pynko / Discrete Mathematics 232 (2001) These equalities collectively imply a This contradiction shows that the four pairs {e c; 0 ;e c; 1 }, where c BD 4, are pairwise disjoint. As A is not singular, by Proposition 3.6 and by the quasi-identity (9) we have e ; 0 e ; 1. Ife 0; 0 was equal to e 0; 1, the set {e ; 0 ;e ; 1 ;e 0; 0 ;e ; 1 ;e 1; 1 } would be a pentagon sublattice ofthe { ; }-reduct L of A, what would contradict distributivity of L. Hence e 0; 0 e 0; 1. Applying the to both inequalities proved and taking into account its injectivity, we also get e ; 0 e ; 1 and e 1; 0 e 1; 1. Thus, e is an embedding of BD 4 BD 2 into A. Therefore BD 4 BD 2 K, sok = I(BD 4 BD 2 ). 2. K * I(BD 4 BD 2 ). Take any A K\I(BD 4 BD 2 ). Then, by Proposition 3.6, A is not singular and there is some a A such that a = a. Consider the mapping e from BD 4 to A dened by e1:= ; e := a; e0:= ; e :=@ a: Using (4), (5) and (6), it is straightforward to check that e hom(bd 4 ; A). As A is not singular, e0 e1, so ker e BD 4 BD 4. On the other hand, ker e is a congruence of BD 4. Recall that BD 4 is simple (for D 4 is simple). Hence e is injective. Thus, e is an embedding of BD 4 into A. Therefore BD 4 K. By Proposition 3.1 this yields K = DMBA: Remark that the lattice ofimplicational classes ofde Morgan Boolean algebras is dierent from those of De Morgan lattices and De Morgan algebras. As it has been proved in [10, Theorem 4:8], the implicational classes ofde Morgan lattices form an eight-element non-chain lattice. Finally, according to [1], there are 2 ℵ0 quasivarieties (not saying about implicational classes) ofde Morgan algebras. Acknowledgements The author is indebted to the anonymous referee for the comments, corrections and suggestions that have helped the author improve the rst version ofthe paper. References [1] M.E. Adams, H.A. Priestley, De Morgan algebras are universal, Discrete Math. 66 (1987) [2] R. Balbes, P. Dwinger, Distributive lattices, University ofmissouri Press, Columbia, MO, [3] N.D. Belnap, A useful four-valued logic, in: J.M. Dunn, G. Epstein (Eds.), Modern uses of Multiple-valued Logic, Reidel, Dordrecht, 1977, pp [4] J.A. Kalman, Lattices with involution, Trans. Amer. Math. Soc. 87 (1958) [5] A.I. Mal cev, Algebraic Systems, Springer, New York, [6] G.C. Moisil, Recherches sur l algebre de la logique, Ann. Sci. Univ. Jassy 22 (1935)

8 66 A.P. Pynko / Discrete Mathematics 232 (2001) [7] A.P. Pynko, Characterizing Belnap s logic via De Morgan s laws, Math. Logic Quart. 41 (1995) [8] A.P. Pynko, Functional completeness and axiomatizability within Belnap s four-valued logic and its expansions, J. Appl. Non-Classical Logics 9 (1999) [9] A.P. Pynko, Denitional equivalence and algebraizability ofgeneralized logical systems, Ann. Pure Appl. Logic 98 (1999) [10] A.P. Pynko, Implicational classes ofde Morgan lattices, Discrete Math. 205 (1999)